name: flask
url: https://github.com/pallets/flask
# Pinned to the latest test_commit SHA so the snapshot is deterministic and
# every test_commit below is reachable as an ancestor.
commit: a29f88ce6f2f9843bd6fcbbfce1390a2071965d6
language: python
size_category: small

test_commits:
  - sha: fbb6f0bc4c60a0bada0e03c3480d0ccf30a3c1df
    description: "all teardown callbacks are called despite errors"
    changed_files: 10
  - sha: a29f88ce6f2f9843bd6fcbbfce1390a2071965d6
    description: "document that headers must be set before streaming"
    changed_files: 4

entry_points:
  - "src/flask/app.py::Flask.wsgi_app"
  - "src/flask/sansio/app.py::App.add_url_rule"

search_queries:
  - query: "Flask wsgi"
    expected: "src/flask/app.py::Flask"
  - query: "AppContext globals"
    expected: "src/flask/ctx.py::AppContext"
  - query: "create logger"
    expected: "src/flask/logging.py::create_logger"

# Multi-hop retrieval tasks (semantic_search → query_graph one-hop)
# See docs/REPRODUCING.md for the schema.
multi_hop_tasks:
  - id: flask-dispatch-callers
    nl_query: "Where Flask dispatches HTTP requests"
    anchor_qualified_suffix: "src/flask/app.py::flask.dispatch_request"
    traversal_pattern: callers_of
    expected_neighbor_names: ["full_dispatch_request"]
    k: 10
  - id: flask-exception-callers
    nl_query: "Where Flask handles uncaught exceptions"
    anchor_qualified_suffix: "src/flask/app.py::flask.handle_exception"
    traversal_pattern: callers_of
    expected_neighbor_names: ["wsgi_app"]
    k: 10

# Questions for the agent_baseline benchmark (pure-python grep top-k vs graph
# query). See docs/REPRODUCING.md for the methodology.
agent_questions:
  - "How does dispatch_request route an incoming HTTP request"
  - "Where is the AppContext pushed and popped"
  - "How does create_logger configure application logging"